Multi-language package manager detection and OpenCode agent framework. Auto-detects npm, yarn, pnpm, poetry, cargo, and 40+ package managers across 16 programming languages.
npm install super-opencodeA comprehensive agent framework and installer for the OpenCode CLI ecosystem, designed to streamline development workflows with specialized agents, reusable skills, and structured command patterns.
Super-OpenCode provides a structured framework for AI-assisted software development, featuring:
- Specialized Agent Personas: Pre-configured agents for different domains (Backend, Frontend, Architecture, Security, QA, etc.)
- Reusable Skills: Modular capabilities for confidence checking, self-correction, debugging, and more
- Slash Commands: Pre-built workflows for common development tasks
- MCP Integration: Built-in support for Model Context Protocol servers
- Quality Standards: Enforced best practices and documentation standards
- Node.js 18+
- npm, pnpm, or yarn
``bash`
npm install -g super-opencode
Then run the interactive installer:
`bash`
super-opencode
The installer will guide you through:
1. Choosing installation scope (global or project)
2. Selecting components to install (agents, commands, skills, core)
3. Configuring optional MCP servers (Context7, Serena, Tavily, etc.)
After npm install, the framework files will be available at:
Global Installation:
- Linux/macOS: ~/.config/opencode/
- Windows:
Project Installation:
- All Platforms: Files are copied to the current working directory under .opencode/ and AGENTS.md
Super-OpenCode includes specialized agents for different domains:
| Agent | Purpose |
|-------|---------|
| pm-agent | Project orchestration, PDCA cycles, documentation |
| architect | System design, architecture, technical strategy |
| backend | APIs, databases, server-side logic |
| frontend | UI/UX, components, styling |
| security | Security review, threat modeling |
| quality | Testing, code review |
| researcher | Deep research, fact-checking |
| writer | Technical documentation |
| reviewer | Code review, quality assurance |
| optimizer | Performance optimization |
Reusable capabilities that agents can invoke:
- confidence-check: Pre-execution risk assessment
- self-check: Post-implementation validation
- security-audit: OWASP Top 10 vulnerability detection
- reflexion: Post-action analysis and learning
- debug-protocol: Root cause analysis workflow
- simplification: Complexity reduction
- package-manager: Multi-language package manager detection (npm, yarn, pnpm, poetry, cargo, and 40+ more)
Pre-built workflows for common tasks:
- /soc-implement - Code implementation with best practices/soc-design
- - System design and architecture/soc-test
- - Test generation and execution/soc-research
- - Deep research and documentation/soc-review
- - Code review and quality checks/soc-brainstorm
- - Idea generation and problem-solving/soc-analyze
- - Codebase analysis/soc-cleanup
- - Code cleanup and refactoring/soc-explain
- - Code explanation/soc-workflow
- - PDCA workflow management/soc-pm
- - Project management tasks/soc-git
- - Git operations/soc-improve
- - Continuous improvement/soc-help
- - Help and documentation
Super-OpenCode includes a powerful multi-language package manager detection system that automatically identifies the correct package manager for any project:
Supported Languages: JavaScript/TypeScript, Python, Go, Rust, Java, Ruby, PHP, .NET, Elixir, Haskell, C/C++, Swift, Scala, Clojure, Julia, R
Quick Usage:
`bashAfter installing super-opencode
detect-pm
Example Output:
json
{
"languages": [
{
"language": "javascript",
"package_manager": "pnpm",
"confidence": 95,
"commands": {
"install": "pnpm install --frozen-lockfile",
"add": "pnpm add"
}
}
]
}
`CI/CD Integration:
yaml
- name: Detect Package Manager
run: |
PM=$(npx detect-pm --recommend)
$PM install
`$3
The framework supports configuration of various MCP servers:
Recommended (Core):
- Context7 - Official documentation lookup
- Serena - Codebase analysis and navigation
- Tavily Search - Web search for research
- Filesystem - File system access
- Sequential Thinking - Multi-step reasoning
Optional:
- GitHub - GitHub API integration
- SQLite - Database operations
- Chrome DevTools - Browser debugging
- Playwright - Browser automation
Project Structure
super-opencode/
├── src/
│ └── cli.ts # Main installer CLI
├── .opencode/
│ ├── agents/ # Agent persona definitions
│ │ ├── architect.md
│ │ ├── backend.md
│ │ ├── frontend.md
│ │ └── ...
│ ├── commands/ # Slash command definitions
│ │ ├── soc-implement.md
│ │ ├── soc-design.md
│ │ └── ...
│ └── skills/ # Reusable skill modules
│ ├── confidence-check/
│ ├── self-check/
│ ├── security-audit/
│ └── ...
├── AGENTS.md # Core principles and guidelines
├── package.json
├── tsconfig.json
└── biome.json # Linting configuration
Usage
$3
1. Read AGENTS.md - This file contains core principles, development workflows, and quality standards
2. Configure MCP Servers - Set up the MCP servers you need for your workflow
3. Use Slash Commands - Invoke the appropriate command for your task
$3
#### Implement a Feature
bash
/soc-implement "User Authentication" --agent backend
`#### Design a System
bash
/soc-design "Microservices Architecture"
`#### Code Review
bash
/soc-review
`#### Research
bash
/soc-research "Best practices for React state management"
`Development Workflow
Super-OpenCode enforces a structured development approach:
$3
- Always verify with official sources
- Use context7 MCP for documentation lookup
- Check existing code before implementing
- Never guess or make assumptions
$3
- Check confidence before starting work
- ≥90%: Proceed with implementation
- 70-89%: Investigate more, present alternatives
- <70%: STOP - ask questions, gather context
$3
- Use Wave → Checkpoint → Wave pattern
- Batch read operations together
- Analyze together before editing
- 3.5x faster than sequential execution
$3
When errors occur:
1. STOP - Don't retry immediately
2. INVESTIGATE - Research root cause
3. HYPOTHESIZE - Form theory with evidence
4. REDESIGN - New approach
5. EXECUTE - Implement based on understanding
6. LEARN - Document for prevention
Configuration
$3
After installation, configuration is stored in platform-specific locations:
Global Configuration:
- Linux:
(follows XDG Base Directory Specification)
- macOS: ~/Library/Application Support/opencode/opencode.json
- Windows: (typically C:\Users\)Project Configuration:
- All Platforms:
./opencode.json in your project rootEnvironment Variables:
-
OPENCODE_CONFIG_DIR: Override the global config directory
- OPENCODE_CONFIG: Override the full path to the config file
- OPENCODE_CONFIG_CONTENT: Provide inline JSON configuration$3
When installing Super-OpenCode globally:
- Linux:
~/.config/opencode/ (follows XDG Base Directory Specification)
- macOS: ~/.config/opencode/ (same as Linux for consistency with OpenCode)
- Windows: %APPDATA%\opencode\ (typically C:\Users\)Note: Framework files (agents, commands, skills) are stored alongside the OpenCode config file for easy access.
Project Installation:
- All Platforms: Current working directory (where you run the installer)
$3
MCP servers are configured in the
mcp section of your config. According to the OpenCode MCP documentation:`json
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"context7": {
"type": "local",
"command": ["npx", "-y", "@upstash/context7-mcp"],
"enabled": true
},
"filesystem": {
"type": "local",
"command": [
"npx",
"-y",
"@modelcontextprotocol/server-filesystem",
"/path/to/project"
]
},
"tavily": {
"type": "local",
"command": ["npx", "-y", "tavily-mcp@latest"],
"environment": {
"TAVILY_API_KEY": "your-api-key"
}
}
}
}
`Note: MCP servers can be enabled/disabled per agent by modifying the agent configuration. For more details, see the official OpenCode MCP docs.
Quality Standards
$3
- All public functions require docstrings
- Use type hints where supported
- Follow existing project patterns
- Include usage examples for complex functions
$3
- Current with "Last Verified" dates
- Minimal but necessary information
- Clear with concrete examples
- Practical and copy-paste ready
$3
- Write tests for new functionality
- Aim for >80% code coverage
- Include edge cases and error conditions
- Run full test suite before major changes
Building and Development
bash
Install dependencies
npm installBuild
npm run buildLint
npm run lintFormat
npm run format
``Contributions are welcome! Please follow these guidelines:
1. Read AGENTS.md to understand the framework philosophy
2. Follow existing code patterns and quality standards
3. Add tests for new functionality
4. Update documentation as needed
5. Submit pull requests with clear descriptions
MIT License - See LICENSE file for details.
Created by lst97
- OpenCode CLI
- Model Context Protocol
- MCP Server Registry
---
Built with ❤️ for the AI-assisted development community